How saketh070706/code_storm works
CODE_STORM is a full-stack GenAI platform whose primary documented workflow is a browser-based enterprise workspace: users manage policies, agents, and sources, then submit governed actions for authorization, review, execution, and audit. The frontend also retains a legacy copilot for chat, document retrieval, and structured extraction. Workspace and copilot persistence have distinct implementations; the README additionally documents vector search and multi-provider LLM resilience. Some subsystem links below rely on the README or SOURCE INDEX; unsampled implementation details are not inferred.
Main entry points: api.py Workspace API · runtime.py Action runtime · chat.py Copilot routes · rag.py Knowledge retrieval
Components 27
Browser workspace
- Workspace UI
frontend/src/App.jsx - Workspace views
frontend/src/workspace/Registry.jsx - Workspace state
frontend/src/workspace/useWorkspace.js - API transport
frontend/src/workspaceClient.js
Backend APIs
- Workspace API
backend/workspace/api.py - Copilot routes
backend/api/routes/chat.py - Extraction API
backend/api/routes/extraction.py
Governance runtime
- Action governance
backend/core/governor_service.py - Action runtime
backend/workspace/runtime.py - Policy rules
backend/workspace/rules.py - Identity and roles
backend/workspace/identity.py - Workflow compiler
backend/workspace/compiler.py - Audit trail
backend/workspace/audit.py
Copilot and retrieval
- Source registry
backend/workspace/sources.py - Knowledge retrieval
backend/core/rag.py - Structured extraction
backend/core/extraction.py - Prompt safety
backend/core/safety_scaffold.py - LLM cascade
backend/core/llm_client.py
Persistence and providers
- Workspace database
backend/workspace/models.py - Governor store
backend/core/governor_store.py - Audit database
backend/core/audit_store.py - Vector search
backend/integrations/pinecone_client.py - Embeddings
backend/integrations/embeddings.py
Other components
- User
- PostgreSQL
- Pinecone
- LLM providers
Connections 31
- User → Workspace UI: uses
README.md - Workspace UI → Workspace state: loads state
frontend/src/App.jsx - Workspace UI → Workspace views: renders views
frontend/src/App.jsx - Workspace state → API transport: uses transport
frontend/src/workspace/useWorkspace.js - API transport → Workspace API: REST requests
README.md - Workspace API → Action runtime: dispatches actions
backend/workspace/api.py - Workspace API → Policy rules: uses rules
backend/workspace/api.py - Workspace API → Identity and roles: uses identity
backend/workspace/api.py - Workspace API → Workflow compiler: uses compiler
backend/workspace/api.py - Workspace API → Source registry: uses sources
backend/workspace/api.py - Workspace API → Knowledge retrieval: uses retrieval
backend/workspace/api.py - Workspace API → Prompt safety: uses safety
backend/workspace/api.py - Workspace API → LLM cascade: uses LLM
backend/workspace/api.py - Action runtime → Policy rules: evaluates policies
backend/workspace/runtime.py - Action runtime → Identity and roles: checks identity
backend/workspace/runtime.py - Action runtime → Audit trail: records actions
backend/workspace/runtime.py - Action runtime → Action governance: uses governor
backend/workspace/runtime.py - Audit trail → Audit database: persists events
backend/workspace/audit.py - Action governance → Governor store: stores decisions
backend/core/governor_service.py - Workspace API → Workspace database: persists workspace
backend/workspace/api.py - Workspace database → PostgreSQL: connects to
README.md - Copilot routes → Knowledge retrieval: retrieves context
backend/api/routes/chat.py - Copilot routes → Prompt safety: checks prompts
backend/api/routes/chat.py - Copilot routes → LLM cascade: requests completion
backend/api/routes/chat.py - Copilot routes → Workspace database: persists conversations
backend/api/routes/chat.py - Extraction API → Structured extraction: runs extraction
- Structured extraction → LLM cascade: requests completion
- Knowledge retrieval → Embeddings: creates embeddings
backend/core/rag.py - Knowledge retrieval → Vector search: queries vectors
backend/core/rag.py - Vector search → Pinecone: searches index
README.md - LLM cascade → LLM providers: cascades providers
backend/core/llm_client.py